IT WOULDN'T BE SUMMER if an AMD takeover rumour wasn't doing the rounds and yesterday Patrick Wang of Evercore Investments claimed that Qualcomm and Samsung are circling as possible suitors for the chip designer.
AMD takeover rumours are not new, they go back decades, but the firm has always managed to keep its identity even while shedding its fab business. While Wang's claim isn't extraordinary, it's hard to see what Qualcomm or Samsung might gain from buying up AMD.
Both Qualcomm and Samsung are high-flying chip vendors with strong balance sheets and steadily increasing market shares, which AMD hasn't been able to say for a long time. With AMD having sold off its fabs, the only major assets it has are an x86 licence from Intel and its own AMD x86-64 chip architecture, which it licenses to Intel.
AMD's x86 licence doesn't hold quite the same value as it did a few years back, with ARM chips dominating the embedded market and starting to move into the server market. With Microsoft porting Windows 8 to the ARM chip architecture, there's very little to stop full-fledged ARM laptops from running Windows. The fact is that x86 is no longer the be-all and end-all for chip vendors.
Nvidia, AMD's traditional competitor, is an example of this. After Intel repeatedly denied it an x86 licence, Nvidia has reinvented itself by focusing on its ARM-based Tegra system-on-chip processor and Tesla high performance computing boards and, more importantly, posting profits.
Where ARM is lagging behind is by not having a true 64-bit architecture, however the firm has said that it is not far away. It is also becoming clear that ARM-based servers will pick up a not-inconsequential chunk of the server market, overtaking the lacklustre market penetration of AMD's Opteron processors.
Then there is AMD's track record of execution. As a chip design outfit, AMD has produced some very interesting designs, the most recent of which have been Bulldozer and Brazos, but while they are interesting and innovative, the performance of both architectures has fallen short.
AMD's disconnect between what is a good design on paper and producing a product that can compete with its rivals in the real world is a very worrying one for anyone looking to buy the firm, because it signals deeply rooted problems in business processes and practices. Such major competitiveness problems cannot be fixed overnight and in some cases cannot be fixed at all.
So what does AMD have to offer Qualcomm and Samsung? The firm's graphics division, which is still essentially ATI, brings in a fair amount of business and it obviously has a portfolio of important graphics patents acquired over decades.
However Qualcomm already nabbed AMD's graphics CTO Eric Demers earlier this year and it too has a library of graphics technologies from its development of the Adreno GPU. And while AMD's current Graphics Core Next architecture is very impressive in the desktop and mobile markets, that doesn't automatically mean it will provide competitive performance if it is scaled down into a GPU core that has a power budget of just a couple of watts.
Given that Qualcomm and AMD both rely heavily on TSMC for stamping out chips, one thing Qualcomm - not so much Samsung - would gain from buying AMD is immediate access to extra capacity at TSMC. Qualcomm raised some eyebrows when it publicly chided TSMC for its 28nm process yield problems and said it would start looking elsewhere for 28nm fab capacity.
Qualcomm's tough talk was hard for it to back up with action because TSMC and its 28nm process node are maturing and altering designs for other process nodes at other fabs can lead to poor yields. With AMD's Southern Islands GPUs effectively being the high volume launch of TSMC's 28nm process node, a Qualcomm buyout of AMD would give it increased buying power and perhaps even put it at the head of the queue at a fab company it simply has to work with to survive.
Aside from AMD's capacity at TSMC, it is hard to see what Qualcomm could gain by acquiring the company. As for Samsung, it already has many fabs and while it has deeper pockets, the most obvious reason for it to spend a few billion on AMD would be to pick up a library of graphics patents to act as a shield against patent infringement claims. µ